search

nmoehrle / mvs-texturing

Algorithm to texture 3D reconstructions from multi-view stereo images
Other
933 stars 328 forks source link

tying cameras to mesh face #75

Closed dakotabenjamin closed 6 years ago

dakotabenjamin commented 7 years ago

We received an interesting feature request over in OpenDroneMap: https://github.com/OpenDroneMap/OpenDroneMap/issues/557

The question is whether we can extract the camera information from each face?

nmoehrle commented 7 years ago

Sorry I somehow lost track of this issue.

There is currently no way to tell which cameras contributed to the reconstruction of a face (talking about geometry) if you use the MVE pipeline. The information gets lost once you export the depth maps into a pset (scene2pset).

If we talk about texture it becomes a little easier since we only use one image to texture a face you, for debug purposes you can visualize this mapping with the --view_selection_model as you probably know. The mapping gets lost once we combine texture patches to texture atlases. If you want to contain this information in a obj you would have to create one texture atlas per input image.